Department of Defense Legacy Resource Management Program · MacKenzie et al. (2002) that accounts...
Transcript of Department of Defense Legacy Resource Management Program · MacKenzie et al. (2002) that accounts...
Department of Defense Legacy Resource Management Program
PROJECT NUMBER (10-343)
Le Conte’s Thrasher (Toxostoma lecontei) Occupancy and Distribution:
Barry M. Goldwater Range and Yuma Proving Ground in Southwestern Arizona
Scott T. Blackman AZ Game and Fish Dept.
September 2012
PUBLIC RELEASE STATEMENT (OPTIONAL) (arial 12 pt)
ii
Le Conte’s Thrasher (Toxostoma lecontei) Occupancy and Distribution: Barry M. Goldwater Range and Yuma Proving Ground in Southwestern Arizona
Scott T. Blackman
Shawn F. Lowery
Joel M. Diamond, PhD
Arizona Game and Fish Department
5000 West Carefree Highway
Phoenix, AZ 85086
September 2012
iii
RECOMMENDED CITATION
Blackman, S. T., S. F. Lowery, and J. M. Diamond. 2012. Le Conte’s Thrasher (Toxostoma
lecontei) Occupancy and Distribution: Barry M. Goldwater Range and Yuma Proving Ground in
Southwestern Arizona. Research Branch, Arizona Game and Fish Department.
ACKNOWLEDGMENTS
Funding was provided by the Department of Defense Legacy Resource Management Program,
Project #10-343. Special thanks to Dennis Abbate, Chris Bertrand, Will Carrol, Josh Ernst,
Hillary Hoffman, Ryan Mann, Ronald Mixon, Eduardo Moreno, and Angela Stingelin for field
support, and to John Arnett of 56th
Range Management Office, Luke Air Force Base, AZ, for
field support and reviewing this document. Vince Frary provided critical insight into the
occupancy modeling framework. This project is also indebted to Ray Schweinsburg, Mike
Ingraldi and Renee Wilcox for valuable logistical and administrative assistance.
The Arizona Game and Fish Department prohibits discrimination on the basis of race, color,
sex, national origin, age and disability in its programs and activities. If anyone believes they
have been discriminated against in any of AGFD’s programs or activities, including its
employment practices, the individual may file a complaint alleging discrimination directly with
AGFD Deputy Director, 5000 W. Carefree Highway, Phx, AZ 85023, (623) 236-3000 or US Fish
and Wildlife Service, 4040 N. Fairfax Dr., Ste. 130, Arlington, VA 22203.
Persons with a disability may request a reasonable accommodation, such as a sign language
interpreter, or this document in an alternative format, by contacting the AGFD Deputy Director,
5000 W. Carefree Highway, Phx, AZ 85023, (623) 236-3000 or by calling TTY at 1-800-367-
8939. Requests should be made as early as possible to allow sufficient time to arrange for
accommodation.
iv
TABLE OF CONTENTS
INTRODUCTION..........................................................................................................................1
GOALS AND OBJECTIVES........................................................................................................2
STUDY AREA ................................................................................................................................2
Barry M. Goldwater Range East and West ..........................................................................2
Yuma Proving Ground .........................................................................................................3
METHODS .....................................................................................................................................4
Prediction of Occurrence Model ..........................................................................................4
Survey Methodology ............................................................................................................5
Habitat and Landscape Data Collection ...............................................................................6
Occupancy Modeling ...........................................................................................................6
RESULTS. ......................................................................................................................................8
Prediction of Occurrence Model ..........................................................................................8
Call-broadcast Surveys ........................................................................................................9
Surveys on YPG ...........................................................................................................9
Surveys on BMGR East .............................................................................................11
Surveys on BMGR West ............................................................................................11
Perch Locations ..................................................................................................................11
Nest Locations ...................................................................................................................11
Occupancy Estimation .......................................................................................................11
Soil Context .......................................................................................................................13
Avian Community ..............................................................................................................13
DISCUSSION ...............................................................................................................................16
MANAGEMENT AND RESEARCH PRIORITIES ................................................................20
REFERENCES .............................................................................................................................23
v
TABLES
Table 1. Candidate set of occupancy models applied to Le Conte’s thrasher habitat data
gathered during repeated surveys on DoD lands in southwestern Arizona.
Estimated parameters include: i = the probability that a species is present at site
i, and pit = the probability that a species is detected at site i during visit t ......................8
Table 2. Number of Le Conte’s thrasher survey sites and perch locations for five Prediction
of Occurrence Model Classes.. ........................................................................................9
Table 3. Mixed occupancy models for covariates supported by the Le Conte’s thrasher
occupancy data compared to the global model, presented with Akaike Information
Criteria (AIC) values, ∆AIC, AIC weight, and likelihood. ...........................................10
Table 4. Mixed model logistic regression for covariates supported by the Le Conte’s thrasher
occupancy data compared to the global model, presented with Akaike Information
Criteria (AIC) values, ∆AIC, AIC weight, and likelihood ...........................................14
Table 5. Untransformed parameter estimates and standard errors (adjusted using variance
inflation factor of 2.45) in most supported Le Conte’s thrasher occupancy model. .....14
Table 6. Le Conte’s thrasher perch location and non-detection locations within respective
NRCS soil map units .....................................................................................................15
Table 7. Le Conte’s thrasher perch location and non-detection locations within respective
NRCS soil associations .................................................................................................15
FIGURES
Figure 1. Locations of Le Conte’s thrasher surveys during 2011 at YPG, BMGR East and
BMGR West. ...................................................................................................................4
Figure 2. Schematic of parallel transects with call-broadcast survey points conducted by two
surveyors walking in opposite directions. The middle 2 points are 800 m apart and
are centered about a randomly-generated point. Other points on each transect are
400 m apart. Transects are 1 km apart ............................................................................6
Figure 3. Plots where Le Conte’s thrashers were and were not detected during surveys at
YPG, BMGR East, and BMGR West during 2011 .......................................................12
APPENDIX
Appendix 1. Randomly generated centers of Le Conte’s thrasher survey plots ...........................25
1
INTRODUCTION
Le Conte’s thrasher (Toxostoma lecontei; hereafter LCTH) is an uncommon permanent
resident of sparsely vegetated landscapes within the San Joaquin Valley, Kern River
basin, Owens Valley, Mojave Desert, and the Lower Colorado River Valley subdivision
of the Sonoran Desert biotic community in the southwestern United States (Sheppard
1996, Corman and Wise-Gervais 2005). Nesting occurs from January through May and,
rarely, into early June (Sheppard 1970, Sheppard 1996, Corman and Wise-Gervais 2005).
This species is listed as a Bird of Conservation Concern by US Fish and Wildlife Service
(USFWS), and as a Wildlife Species of Concern by the Arizona and California Game and
Fish Departments (Latta et al. 1999, CalPIF 2006). Within Arizona, the densest
concentrations of LCTH occur on the Cabeza Prieta National Wildlife Refuge and the
Barry M. Goldwater Range (BMGR) (Corman and Wise-Gervais 2005). Density
estimates in California have ranged from 0.2-7.3 pairs/km² (CalPIF 2006). Populations
have declined in some areas including California’s San Joaquin Valley (CalPIF 2006,
CMSHCP 2007) and in Arizona where agriculture and urban development have impacted
this thrasher’s habitat.
The Department of Defense (DoD) manages large tracts of Sonoran Desert and
correspondingly plays a major role in the conservation of this ecoregion (Marshall 2000).
The BMGR encompasses 1,733,921 acres (701,718 hectares, 2,709 square miles, or 7,016
square kilometers) and is jointly managed by the U.S. Air Force and U.S. Marine Corps
to train military aircrews for air combat missions (BMGR 2007a). The Yuma Proving
Ground (YPG) contains approximately 3,450 km² of Sonoran Desert in La Paz and Yuma
counties and is currently used for testing training equipment and personnel in the harsh
desert environment (Figure 1). As federal land managers, BMGR and YPG personnel
comply with the Sikes Act and the Endangered Species Act as part of installation
operations. Information on sensitive, threatened and endangered species that potentially
occur on BMGR and YPG is needed to make military planning compatible with sensitive
species management.
Natural resource monitoring and management at BMGR and YPG is guided by Integrated
Natural Resources Management Plans (BMGR 2007a, USYPG 1998) and Inventory and
Monitoring Plans (BMGR 2007b, Villarreal et al. 2011). Military activities on BMGR
and YPG such as ground-based training and heavy equipment maneuvers involving
wheeled and tracked vehicles may negatively impact LCTH. Other military activities
that cause moderate to high levels of disturbance to soils and vegetation (e.g., explosive
ordnance clearance areas, munitions impact areas) can also threaten LCTH on BMGR
and YPG if activities occur within known breeding areas or potential habitat.
The proportion of area occupied (PAO) is a popular alternative to abundance estimation
in wildlife monitoring programs because, unlike abundance estimates, PAO metrics
incorporate the detection probability of each species (Bailey et al. 2004, MacKenzie and
Royle 2005). If the detection probability for a species is not incorporated into occupancy
estimates, a naïve count of the area (the number of sites occupied by the species divided
2
by the total number of sites surveyed) will underestimate the actual site occupancy
(MacKenzie et al. 2002, MacKenzie et al. 2003, Tyre et al. 2003, MacKenzie and Nichols
2004). PAO estimates are calculated using the likelihood-based approach described by
MacKenzie et al. (2002) that accounts for species or individuals present but undetected
during surveys.
GOALS AND OBJECTIVES
The goals of this project were to elucidate the PAO of LCTH on BMGR and YPG in
southwestern Arizona during the 2011 breeding season. Results from the 2011 LCTH
surveys will provide useful information about the distribution of this species on BMGR
and YPG and highlight potential habitat attributes that facilitate this species site-specific
occupancy. Additionally, we developed a pattern recognition model to predict highly
probable LCTH habitat. This will assist BMGR and YPG to manage LCTH for long-
term sustainability across these military installations. Our objectives were as follows:
1) Develop a LCTH Prediction of Occurrence Model;
2) Survey for the presence of LCTH within BMGR and YPG and relate site-specific
occupancy to habitat attributes; and
3) Determine Proportion of Area Occupied (PAO) for LCTH on BMGR and YPG.
STUDY AREA
Barry M. Goldwater Range East and West
The Barry M. Goldwater Range is co-managed by the U.S. Air Force (USAF) and the
U.S. Marine Corps (USMC). The land-management authority for the eastern 1.1 million
acres is the 56th
Range Management Office (56 RMO) at Luke Air Force Base, Phoenix,
AZ. The western portion, more than 600,000 acres, is managed by the Range
Management Department at Marine Corps Air Station Yuma in Yuma, AZ. The Range
occupies portions of Pima, Maricopa and Yuma counties, from the City of Yuma to
several miles East of Gila Bend, Arizona, and totals approximately 7,066 km2 (Figure 1).
The Range is bounded to the south by Mexico and Cabeza Prieta National Wildlife
Refuge, to the north by Interstate-8 and a mix of private and public properties, and to the
east by the Tohono O’odham Nation and Bureau of Land Management lands.
Elevations at BMGR range from below 200 ft at western portions of the Range to 3,700 ft
in the Sand Tank Mountains at the eastern border (BMGR 2007a). Temperatures on
BMGR can range from below 0° C (rare) to 49° C, with a range-wide average annual
rainfall of approximately 5 inches (BMGR 2007a).
The Lower Colorado River subdivision of the Sonoran Desert is the predominating
vegetative community and is characterized by drought-tolerant plant species such as
creosote (Larrea tridentata), bursage (Ambrosia spp.), paloverde (Parkinsonia spp.) and
cacti (e.g., Cylindropuntia spp. and Carnegiea gigantea) (Brown 1994, Marshall et al.
3
2000). The broad, flat and sparsely vegetated desert plains of BMGR are dissected by
incised washes characterized by paloverde, ironwood (Olneya tesota), smoketree
(Psorothamnus spinosus), catclaw acacia (Acacia greggii), mesquite (Prosopis spp.),
ocotillo (Fouquieria splendens) and other shrubs. The Arizona Upland Subdivision of
the Sonoran Desert occurs on elevated hills and mountain slopes of BMGR East,
primarily east of State Route 85. Because LCTH does not inhabit the Upland
Subdivision, we do not provide a detailed description of this subdivision.
Yuma Proving Ground
The Yuma Proving Ground is managed by the U.S. Army. YPG occupies portions of La
Paz and Yuma counties near Yuma, Arizona, and totals approximately 3,450 km² (Figure
1). Kofa National Wildlife Refuge and YPG share a 58-mile long boundary (USDI
1996). The elevation at YPG ranges from sea level to 878m. Average temperatures
range from 16° C (December) to 30° C (July) (Atmospheric Sciences Laboratory, YPG
Central Meteorological Observatory), with average annual rainfall of approximately 8.8
cm.
The prevalent vegetative community on YPG is the Lower Colorado River subdivision of
the Sonoran Desert, described above. As at BMGR, the broad, flat plains of YPG are
dissected by numerous incised washes. The elevated hills and mountain slopes at YPG
are within the Sonoran Desert’s Arizona Upland Subdivision, where plants such as
beargrass, cacti and agave occur.
4
Figure 1. Occupancy surveys for Le Conte’s thrasher during 2011 were located at YPG, BMGR East and
BMGR West.
METHODS
Prediction of Occurrence (PO) Model
We developed a Geographic Information System (GIS) model to predict LCTH
occurrence based on the habitat suitability of the study region and surrounding areas.
Model inputs included vegetative cover (SWReGAP), soil series (Natural Resource
Conservation Service, NRCS), elevation, and previous LCTH detection locations
(Blackman et al. 2010). The GIS model produced a 10-category ranking of potential for
LCTH occurrence throughout the modeled area ranging from category one (least suitable
LCTH habitat) to ten (most suitable LCTH habitat). We omitted land areas classified in
categories 1-5 (least suitable habitat) from further field surveys and analyses because
these areas incorporate large amounts of land cover types known to be unsuitable for
LCTH. We also omitted areas with limited access and hazard areas including bombing
ranges, drop zones and testing (e.g., explosives) ranges.
Using the five best-fitting PO Model categories (categories 6-10), we used ArcMap
(Environmental Research Institute, Redlands, California, USA) to randomly generate
forty (40) points throughout BMGR East, BMGR West and YPG. These forty points
5
became the center of our survey plots, and were at least 3km apart to ensure
independence among LCTH detections. The location of these points was also governed
by limited or restricted access areas on the DoD installations. These restricted areas
included bombing ranges, drop zones and test ranges as they occur on the three
installations. For example, a large portion of YPG on the southern arm (adjoining the
Cibola and Kofa arms) contains numerous large restricted areas where explosives are
tested. These areas were omitted from survey point distribution due to the completely
restricted or extremely limited schedule available for LCTH surveys.
Survey Methodology
Despite inhabiting very sparse landscapes, LCTH can be difficult to detect. These birds
are secretive; the colors of their plumage matches the soil surface, and typically forage on
the ground beneath shrubs and trees unless enticed to a high perch where the bird may
vocalize. Research on this species in the San Joaquin Valley, California, determined that
conducting broadcast surveys (i.e., tape-playback calls) is an effective survey technique,
especially when compared to walking transects where vocal responses are not elicited
(i.e., no broadcasting) (CMHCP 2007).
Survey points were spaced 400 meters apart along transects projecting out from the
center of each randomly generated plot (N=30). Two observers began at the center of
each randomly generated plot and walked in opposing directions (e.g., North/South or
East/West). Broadcast points along each transect were spaced at 400-meter intervals and
both surveyors commenced broadcasting once they had walked 400 meters from the
original random point (Figure 2). After conducting the first point broadcast, each
surveyor then walked 400 meters to the next point. Transects included five points along
one transect and five points along a second transect parallel to and 1,000 m away from
the original transect (Figure 2). Upon completion of the first survey transect, each
surveyor moved 1 km perpendicular to the first transect line to start the second transect
line. The second transects were parallel to the first transect and the direction that the
surveyor chose to begin the second transect was contingent upon the suitability of the
landscape to LCTH occurrence. Double counting was eliminated by skipping broadcast
points directly adjacent to points where LCTH were detected if detected birds began to
follow the observer.
At each broadcast point, surveyors first spent one minute quietly looking and listening for
LCTH. At the conclusion of the first minute, each surveyor broadcast a recording of
LCTH vocalizations for 90 seconds in a direction perpendicular to the transect line,
followed by a 2-minute period of observation. The observer then broadcast the LCTH
vocalizations for another 90 seconds in the direction opposite of the first broadcast
direction and perpendicular to the transect line, followed by another 2 minutes of
observation. If no LCTH were detected, total survey time at each point was 8 minutes. If
a LCTH was detected, the observer stopped the broadcast, spent 15 to 20 minutes
observing the LCTH and recording relevant data (see next paragraph), and then moved to
the next point. When LCTH were detected and if the bird followed the observer, we
reduced the likelihood of double counting (i.e., repeated counts of an individual bird) by
skipping adjacent broadcast points.
6
All surveyors documented the location and tree/shrub species of the perch where each
LCTH was first detected. Perch location was recorded using a hand held Garmin (GPS)
using the NAD 83 datum projected in UTM Zone 11 (western portion of study area) and
12. Perches were marked with flagging for future measurements. We identified and
measured the distance to all birds detected at survey points (Buckland et al. 2001).
Figure 2. Schematic of parallel transects with call-broadcast survey points conducted by two surveyors
walking in opposite directions. The middle 2 points are 800 m apart and are centered about a randomly-
generated point. Other points on each transect are 400 m apart. Transects are 1 km apart.
Habitat and Landscape Data Collection
For all LCTH perches and confirmed nests, we recorded the location, described the perch
or nest substrate, identified the tree or shrub species and estimated the height of the perch
or nest tree. We identified all trees and shrubs within 10 m of all perches, nests, and
locations where LCTH were first detected during surveys. At all locations where we
detected LCTH, and at alternating broadcast stations, we measured habitat characteristics
such as vegetation diversity, proportion of ground cover, percent shrub and tree cover,
and the distances to the nearest tree and ephemeral wash. These data were used as
covariates within the occupancy modeling framework of LCTH across the 3 DoD
installations.
Occupancy Modeling
We used occupancy modeling (MacKenzie et al. 2002) to estimate the occurrence
probability and detectability of LCTH throughout the study area and correlate
presence/absence with covariates within an information-theoretic context (Burnham and
Anderson 2002). Parameters estimated include; ( i ) = the probability that a species is
present at site i, and pit = the probability that a species is detected at site i during visit t.
Selection of survey locations did not require the presence of LCTH, however, survey
locations were randomly generated within boundaries predicted as highly suitable for
LCTH occurrence. Randomization and a lack of specific pre-existing knowledge of
End Transect 2 (B)
End Transect 1 (A)
Start of Surveys (these 2
points are 800 m apart)
1000 m
400 m
7
LCTH site occupancy eliminated site selection bias (MacKenzie and Royle 2005, Collier
et al. 2010).
We developed a priori models, formed on the basis of LCTH biology and life history
strategies, as a foundation for models used for estimating LCTH detection and occupancy
probabilities (MacKenzie et al. 2006). A candidate suite of models contained habitat
(e.g., number of trees, distance to nearest wash, distance to nearest tree, sand and gravel
cover composition) and landscape attributes (e.g., NRCS soil series classification)
associated with LCTH presence (Table 1). We reduced the number of candidate models
by evaluating the influence of survey pass on detection probability while holding
occupancy constant [ψ(.) p(time)]. We then used the most parsimonious model of
detection probability [ψ(.) p(time)] to model the influence of habitat covariates on LCTH
occupancy (Kroll et al. 2007, Hansen et al. 2011).
We used the software program PRESENCE version 4.0 (Hines 2010) to model the
probability of detection and occupancy with habitat and landscape covariates measured at
LCTH detection points and alternating broadcast points across the study areas. LCTH
presence/absence data were analyzed at differing spatial scales to generate an occupancy
spectrum. This multi-scale method ranged from modeling all individual points together,
modeling individual broadcast points consisting of even and odd only analyses (i.e.
alternating broadcast points modeled together), broadcast points pooled with respect to
Transect A and B, and occupancy modeling at the LCTH plot scale.
Akaike’s Information Criterion (AIC) was used to rank the set of considered models in
order of goodness of fit (MacKenzie and Bailey 2004) and compare AIC weights and
∆AIC to assess model uncertainty (Burnham and Anderson 2002). We ranked all
candidate models with respect to AIC values and interpreted the lowest AIC value as the
best model. Models within <2∆AIC of the highest ranked model were considered to be
best supported by the data and competed with the most parsimonious model.
Overdispersion in the data was assessed by testing overall model fit of the global model
by completing 10,000 parametric bootstraps and using the Pearson chi-square statistic to
obtain the variance inflation factor (ĉ) (Burnham and Anderson 2002). Model selection
uncertainty was accounted for by computing untransformed parameter and variance
estimates within the most supported models (Burnham and Anderson 2002). The AIC
weights were summed across covariates represented in the most competitive models
ranking within <2∆AIC of the highest ranked model to assess the relative importance of
the individual covariates.
8
Table 1. Candidate set of occupancy models applied to Le Conte’s thrasher habitat data gathered during
repeated surveys on the DoD lands in southwestern Arizona. Estimated parameters include: i = the
probability that a species is present at site i, and pit = the probability that a species is detected at site i
during visit t.
Occupancy Model Model Description
ψ(.) p(.) Constant occupancy, constant detection
ψ(.) p(t) Constant occupancy, survey pass dependent detection
ψ(Soil) p(t) Soil class dependent occupancy, time dependent detection
ψ(PM) p(t) LCTH Prediction Model dependent occupancy, time dependent
detection
ψ(#Tree) p(t) # tree species dependent occupancy, time dependent detection
ψ(DW*) p(t) Distance to wash dependent occupancy, time dependent detection
ψ(DT*) p(t) Distance to nearest tree dependent occupancy, time dependent
detection
ψ(Sand) p(t) Percent sand composition dependent occupancy, time dependent
detection
ψ(Gravel) p(t) Percent gravel composition dependent occupancy, time dependent
detection *Distance to nearest wash intervals (separate model variables): 0-10m, 10-50m, 50-100m and >100m.
Distance to nearest tree intervals (separate model variables): 10-50m, 50-100m and >100m.
RESULTS
Prediction of Occurrence Model
Le Conte’s thrasher location data consisting of actual perch locations and points where
birds were not detected (Blackman et al. 2010) were modeled with the LCTH Prediction
of Occurrence (PO) model for BMGR and YPG. These data are presented in Table 4
with respect to LCTH PO classes 6-10 as these were the best fitting model classes for
LCTH occurrence. All but class 10 exhibited increases in the ratio between LCTH perch
locations and non-detection locations with respect to increasing predictive power (Table
2). However, the PO Model performed poorly when used as a covariate in occupancy
modeling (Table 3).
9
Table 2. Number of Le Conte’s thrasher survey sites and perch locations for five Prediction of Occurrence
Model Classes.
Prediction of
Occurrence
Model Class
Number of Survey
Sites and Percentage
of Total Sites
Number of LCTH
Perches and Percentage
of Total Perches
Percentage of Sites
with LCTH Perches
Within Each PO Class
6 214 (28) 17 (15) 7.9
7 195 (26) 17 (15) 8.7
8 215 (28) 48 (44) 22.3
9 87 (11) 23 (21) 26.4
10 46 (6) 5 (5) 10.9
Call-broadcast Surveys
We conducted surveys for Le Conte’s thrashers from January to April 2011. Across the
three DoD installations, we detected 183 LCTH at 107 points within 28 plots.
Additionally, ten LCTH were observed incidentally while observers walked between
survey points or en route to surveys; these 10 LCTH were found at plots where LCTH
were detected from established survey points and were not included in occupancy
analyses.
Le Conte’s thrashers bred at our study area during our surveys. We found three active
LCTH nests, and probably detected several male-female pairs. Two LCTH were
simultaneously detected from 48 points within 21 plots, potentially consisting of pairs.
Observers simultaneously detected three LCTH from five points within five different
plots. Because thirteen of the detections consisting of two or more LCTH observations
were made after March 1, at a time when we would expect to observe LCTH pairs and/or
fledglings, this suggests that breeding had occurred or was in progress at these 13
locations.
Surveys at YPG
Because of restricted access and that LCTH are not likely to occur at vast areas of YPG,
we conducted relatively few (n=8, 20% of all surveys) LCTH surveys on YPG. We did
not detect LCTH at all three plots (10, 15 and 32) located in the Cibola Arm of YPG.
Most of the soil surface within the Cibola Arm is desert pavement, a substrate that LCTH
finds unsuitable (Blackman et al. 2010). Likewise, we did not detect LCTH at two of the
plots (22 and 22-2) north of the Tank Mountains on the Kofa Arm where desert pavement
is prevalent. However, we detected LCTH at all three plots (14, 20 and 43) south of the
Palomas Mountains on the Kofa Arm where the soil surface was predominantly softer
sands with relatively less gravel (Appendix 1).
10
Table 3. Mixed occupancy models for covariates supported by the Le Conte’s thrasher occupancy data as
compared to the global model, presented with Akaike Information Criteria (AIC) values, ∆AIC, AIC
weight, and likelihood.
Model AIC ∆AIC AIC w K -2Log
G+(#T)+DW(0-10m) +DT(10-50m) 697.42 0.0 0.3104 6 685.42
G+(#T)+DW(0-10m) +DT(10-50m)+S 697.95 0.53 0.2381 7 683.95
G+(#T)+DW(0-10m) +DT(10-50m)
+DW(10-50m)
699.04 1.62 0.1381 7 685.04
G+(#T)+DW(0-10m) +DT(10-50m) +T>100m 699.16 1.74 0.1300 7 685.16
G+(#T)+DW(0-10m) +DT(10-50m)+S
+ DW(10-50m)
699.70 2.28 0.0993 8 683.70
G+(#T)+DW(0-10m) +DT(10-50m)+S
+DW(10-50m) +T>100m
701.34 3.92 0.0437 9 683.34
G+T(10-50m)
+DW(0-10m)
702.95 5.53 0.0195 5 692.95
G+T(10-50m)+#T 703.68 6.26 0.0136 5 693.68
Global 705.90 8.48 0.0045 13 679.90
G+T(10-50m) 708.14 10.72 0.0015 4 700.14
G+#T+DW(0-10m) 709.23 11.81 0.0008 5 699.23
Gravel (G) 712.03 14.61 0.0002 3 706.03
G+#T 712.76 15.34 0.0001 4 704.76
G+#T+S 712.77 15.35 0.0001 5 702.77
Global-G 715.59 18.17 >0.0001 12 691.59
(#T)+DW(0-10m)
+DT(10-50m)
720.25 22.83 >0.0001 5 710.25
P(T) 739.71 42.29 >0.0001 4 731.71
T(10-50m) 744.09 46.67 >0.0001 3 738.09
11
Surveys at BMGR East
Fourteen plots were randomly distributed across BMGR East (Figure 3). Most surveys
conducted on BMGR East were located in the San Cristobal Valley, east of the Mohawk
Mountains. Consistent with the results of the LCTH surveys we conducted at the San
Cristobal Valley in 2009 (Blackman et al. 2010), during this study we detected LCTH at
53 locations at ten (71%) plots (Appendix 1). We could not access a plot south of
Sentinel and a plot east of SR 85.
Surveys at BMGR West
Our model predicted that much of BMGR West would be suitable LCTH habitat. As a
result, a large proportion (18 of 40, 45%) of our survey plots occurred within BMGR
West even though BMGR West makes up a smaller proportion of our total study area. Of
the eighteen plots at BMGR West, observers detected LCTH at 64 locations within 15 of
the 18 survey plots (Appendix 1). LCTH were not detected at plots 18, 23, and 31
(Figure 3).
Perch Locations
The number of trees within 10 m of LCTH perch locations ranged from 0 (n=56), 1
(n=36), 2 (n=12) and 3 (n=3). If no trees were observed within 10m of LCTH perch
locations (n=56), we recorded the distance to the closest tree as follows: 10-50m (n=23),
50-100m (n=9), and >100m (n=14). The distances from LCTH perch locations to the
nearest wash (of any size) were 0-10m (n=56), 10-50m (n=28), 50-100 (n=10) and
>100m (n=13).
Nest Locations
We found three active nests. We found one tree within 10 m of two nest locations. No
trees were within 10 m of the third nest; this nest was within a cholla cactus. If no trees
were observed within 10 m of LCTH nests, we used the following distances to the nearest
tree: 10-50 m (n=1), 50-100 m (n=0), and >100 m (n=0). The desert wash (of any size)
nearest to the three LCTH nests were 0-10 m (n=2) and >100 m (n=1) away. Nests were
constructed in trees (a paloverde and a mesquite) and shrubs (a cholla cactus). Other
species available for LCTH nest placement included blue paloverde, ironwood, and
crucifixion thorn. We had insufficient data to determine which plants LCTH favor for
nesting.
Occupancy Estimation
The estimated proportion of area occupied (PAO) by LCTH across the three DoD
installations was 0.45 (SE +0.06) and the naïve abundance estimate was 0.14. The
probability of LCTH detection across all survey points was 0.11 (SE +0.02). Occupancy
modeling with only the odd survey points along transects produced an occupancy
probability of 0.60 (SE +0.23) and a probability of detection of 0.09 (SE +0.04).
Occupancy model results using only the even survey points along transects producing an
occupancy probability of 0.35 (SE +0.11) and a probability of detection of 0.14 (SE
+0.05). Pooling all points along transect A produced an occupancy probability of 0.88
(SE +0.40) and a detection probability of 0.07 (SE +0.03). Occupancy models for all
Transect B points produced an occupancy probability of 0.28 (SE +0.08) and a detection
12
probability of 0.17 (SE +0.06). Modeling at the plot scale produced an occupancy
probability of 0.76 (SE +0.08) and a model-averaged detection probability of 0.64 (SE
+0.09).
Figure 3. Plots where Le Conte’s thrashers were and were not detected during surveys at
YPG, BMGR East, and BMGR West during 2011.
Overdispersion was evident in the global occupancy model (√ĉ = 2.45) and the standard
errors were adjusted using the variance inflation factor. The highest ranking occupancy
model contained four covariates: percent gravel composition, total number of trees within
the plot, distance to nearest tree of 10-50 m when trees were not present within the plot
and a nearest wash distance of 0-10 m (Table 3). Three other models were within <2
∆AIC of the most parsimonious model. These models contained the four covariates used
in the highest ranking model and, in order of decreasing importance, percent sand
composition, nearest wash distance of 10-50 m and nearest tree distance of >100 m.
Percent gravel composition contained the highest parameter importance, followed by, in
decreasing order of importance, nearest tree distance of 10-50m, nearest wash distance of
0-10m, and number of trees on plot (Table 3).
13
Model selection uncertainty of the most supported models was fairly high (AICw <0.30).
Therefore, we examined untransformed parameter estimates from covariates included in
the best supported model (Table 4). Untransformed standard errors were high for all of
the parameters included in the most parsimonious model due to the high degree of model
selection uncertainty (Burnham and Anderson 2002). Additionally, we summed the
AICw for covariates represented in the most competitive models ranking within <2∆AIC
of the highest ranked model.
Soil Context
We assigned NRCS soil map units and/or associations to all locations where we detected
LCTH, and to a set of randomly selected points where LCTH were not detected. Because
the NRCS soil map unit GIS layer was not available for YPG, we could not make a direct
comparison between YPG and BMGR based on soil map units. As a result, NRCS soil
associations were used as a surrogate.
The majority of the points (including LCTH detections and the random non-detection
points) were within the Rositas Soil Complex (Rositas sand and Rositas-Ligurta map
units, Table 5). Other prevalent soil map units without LCTH detections were the
Cheroni-Cooledge-Hyder, Gunsight-Hyder-Riverwash, Lomitas-Rock outcrop-Quilotosa,
and Mohall-Pahaka-Valencia (Table 5). The majority of the detection and non-detection
locations for the Soil Association level of detail were contained within the Tremant-
Coolidge-Mohall classification (Table 6). Soil associations without LCTH detections
were the Gunsight-Rillito-Pinal, Laveen-Rillito, and Lithic camborthids-Rock Outcrop-
Lithic Haplagrids (Table 6).
Avian Community
During LCTH surveys we detected the following 63 species: American kestrel, Anna’s
hummingbird, ash-throated flycatcher, bank swallow, Bendire’s thrasher, black-chinned
hummingbird, black-tailed gnatcatcher, black-throated sparrow, blue-gray gnatcatcher,
Bullock’s oriole, burrowing owl, cactus wren, Costa’s hummingbird, common raven,
crissal thrasher, curve-billed thrasher, Eurasian collared-dove, European starling,
Gambel’s quail, gila woodpecker, gilded flicker, golden eagle, great horned owl, greater
roadrunner, Harris’s hawk, hooded oriole, horned lark, house finch, killdeer, ladder-
backed woodpecker, loggerhead shrike, lesser goldfinch, mourning dove, northern
flicker, northern harrier, northern mockingbird, northern rough-winged swallow, orange-
crowned warbler, osprey, phainopepla, red-tailed hawk, rock wren, rufous hummingbird,
sage sparrow, sage thrasher, Say’s phoebe, Scott’s oriole, short-eared owl, Townsend’s
warbler, turkey vulture, verdin, vermilion flycatcher, vesper sparrow, western kingbird,
western meadowlark, western tanager, white-crowned sparrow, white-throated swift,
white-winged dove, Wilson’s warbler, yellow-rumped warbler, and yellow warbler.
14
Table 4. Mixed model logistic regression for covariates supported by the Le Conte’s thrasher occupancy
data as compared to the global model, presented with Akaike Information Criteria (AIC) values, ∆AIC,
AIC weight, and likelihood.
Model AIC ∆AIC AIC w K -2Log
DW 0-10m 754.14 56.72 >0.0001 3 748.14
#Trees 755.03 57.61 >0.0001 3 749.03
Sand 755.19 57.77 >0.0001 3 749.19
DW 10-50m 755.37 57.95 >0.0001 3 749.37
T>100 760.23 62.81 >0.0001 3 754.23
Soil Series
(Torriothents)
761.51 64.09 >0.0001 3 755.51
Table 5. Untransformed parameter estimates and standard errors (adjusted using variance inflation factor
of 2.45) in most supported Le Conte’s thrasher occupancy model.
Model Parameter Est. SE
Gravel -1.2553 2.8803
# trees -1.894 3.4825
DW 0-10m 2.2871 4.1015
DT 10-50m -4.175 5.0737
15
Table 6. Le Conte’s thrasher perch location and non-detection locations within respective soil map units.
Map Unit Name LCTH Perch Locations
(count and percent)
Non-detection Locations
(count and percent)
Rositas-Ligurta 36 35.6 177 27.0
Rositas sand 34 33.7 221 33.7
Torriorthents-
Torrifluvents
8 7.9 74 11.3
Gunsight-Pinamt-
Carrizo
6 5.9 34 5.2
Laposa-Schenco-Rock
outcrop
6 5.9 71 10.8
Wellton loamy sand 4 4.0 13 2.0
Harqua-Tremant 3 3.0 22 3.4
Wellton-Dateland-
Rositas
2 2.0 8 1.2
Antho Sandy Loam 1 1.0 1 0.2
Laposa-Rock outcrop 1 1.0 20 3.1
Cheroni-Cooledge-
Hyder
0 0.0 3 0.5
Gunsight-Hyder-
Riverwash
0 0.0 3 0.5
Lomitas-Rock outcrop-
Quilotosa
0 0.0 7 1.1
Mohall-Pahaka-Valencia 0 0.0 1 0.2
Table 7. Le Conte’s thrasher perch location and non-detection locations in NRCS soil associations.
Map Unit Name LCTH Perch
Locations
Non-detection
Locations
Tremant-Coolidge-Mohall 90 80.4 441 60.8
Harqua-Perryville-Gunsight 11 9.8 149 20.6
Torrifluvents 9 8.0 48 6.6
Supersition-Rositas 2 1.8 14 1.9
Gunsight-Rillito-Pinal 0 0.0 38 5.2
Laveen-Rillito 0 0.0 8 1.1
Lithic camborthids-Rock Outcrop-Lithic
Haplagrids
0 0.0 27 3.7
16
DISCUSSION
LCTH Prediction of Occurrence (PO) Model classes (6-10) were selected for analyses
because these categories predicted the areas with the highest probability of LCTH
occurrence throughout the modeled area. Categories 1-5 incorporated large areas
unsuitable for LCTH occurrence, and were omitted from any analyses. The PO model
performed poorly when used as a covariate in occupancy modeling, ranking well below
the global model. Interestingly, most of the LCTH detections resided in category 8 of the
LCTH PO model despite category 9 and 10 containing a higher LCTH predictive power.
However, category 9 contained the highest proportion of detection/nondetection when
compared to the other four classes. Category 10 had a low detection/nondetection
proportion; however, this class also contained the lowest sample size. Despite the poor
performance of the PO model as a covariate in the occupancy modeling framework, the
model still functioned at the operational level for LCTH occurrence within predictive
classes 6-9. Class 10 may still hold promise for predicting LCTH occurrence but is
difficult to validate because of its rarity in comparison to the other PO classes.
Categories 6-10 of the PO Model can be overlaid into a geo-referenced map in Arcview
providing land managers accurate maps of areas where LCTH are most likely to occur
within their jurisdiction. This could be a powerful conservation tool for land managers,
helping to identify priority LCTH habitat that may exist in proposed footprints for
military activities or development (e.g., alternative energy construction areas).
Surveys conducted in 2009, primarily at the San Cristobal Valley within BMGR East,
verified the purported high abundance of LCTH in this portion of the state (Blackman et
al. 2010). During those surveys, the distribution of LCTH detections were not uniform
across the sampled area and were most concentrated within the center of the valleys
where softer sand predominated and trees were sparse (compared to the mountain
foothills). Correspondingly, in that study, LCTH were not detected at all survey locations
where this species occurrence was predicted by the PO model and where the landscape
appeared suitable.
LCTH surveys in 2011 encompassed a larger area throughout DoD lands in southwestern
Arizona and consisted of 3 survey passes within an occupancy modeling framework. As
expected, these occupancy surveys documented more LCTH locations than surveys
conducted in 2009 exclusively in the San Cristobal Valley with only one survey pass.
Among the three DoD installations we surveyed, most LCTH were detected at BMGR
(East and West), in part because most survey plots were located at BMGR. Within
BMGR, LCTH were not detected at plots close to mountain foothills (i.e., bajadas).
Likewise, of the three areas we surveyed at YPG, LCTH were detected only in the Kofa
arm south of the Palomas Mountains where substrates are softer than at other portions of
YPG. Throughout the study area, areas with a gravelly or desert pavement surface lacked
LCTH.
No single point within the 28 plots where LCTH were detected was occupied by LCTH
during all three survey passes. This demonstrates the general difficulty in detecting
LCTH, particularly late during the breeding season. Thirty-six points had LCTH
17
detections only on the second of three passes (February-March). LCTH were not
detected until the third survey pass at sixteen points (April-May). These detection results
with respect to survey pass explain the low LCTH detection probability and highlight the
significance of its incorporation into population estimation. Additionally, PAO estimates
were higher than naïve occupancy estimates and emphasize that occupancy estimates will
be negatively biased when detection probabilities are not incorporated.
LCTH territories have been reported to be ovate in shape, consisting of 400-450m long
and 200-300m wide (Sheppard 1996). Because survey plots were larger than LCTH
home ranges, plot-scale occupancy models inflated LCTH occupancy and detection
probabilities. Although modeling at the LCTH plot scale produces the highest occupancy
and detection probabilities, this model reduces the amount of data available for LCTH
distribution and can only incorporate landscape-scale covariates. These results
demonstrate how occupancy and detection probabilities can vary depending upon the
scale of the analysis. Modeling at the alternating point (odd and even analysis) scale
effectively increased the point-scale radius to 400 m, as compared to a 200 m radius
comprising the overall broadcast point survey design. Thus, increasing the sampling
distance to 800 m would more accurately portray LCTH home range and ensure
individual point independence, but would sacrifice individual detection locations and
potentially confound raw distribution data. The disparity between occupancy
probabilities of even and odd survey points, and between transect A and transect B,
indicates that relying on either data set independently would omit individual location data
important for mapping LCTH distribution and potentially underestimate occupancy
probabilities. Consequently, we recommend that future survey efforts maintain the
current LCTH survey protocol and analyze presence/absence data at the plot scale.
Habitat data collected during LCTH surveys in the San Cristobal Valley during the 2009
study revealed that two sand cover categories (packed sand and soft sand) represented the
most cover across all LCTH use plots (Blackman et al. 2010). “Packed” and “soft” sand
categories both represented substrates conducive to LCTH ground gleaning and digging.
Additionally, all 8 plots where LCTH were not detected during the 2009 study were
dominated by either hard packed sand or gravel (Blackman et al. 2010). Similar
conclusions can be made about LCTH survey results from this study. LCTH were
unlikely to be detected at plots near mountains where the soil surface has a relatively high
amount of gravel and tree densities are higher than in the lowlands away from the
mountains. We anticipated that gravel composition would have a negative influence on
PAO (i.e., increases in gravel percent composition were inversely proportional to LCTH
occurrence). Gravel composition was the highest ranked individual covariate model and
contained a parameter estimate indicating a negative relationship (Tables 3 and 4).
LCTH appears to avoid areas where the proportion of gravel at the soil surface is above a
certain threshold.
Although the radii sampled around LCTH detection locations were much smaller than
LCTH home ranges, the data collected allows for inferences to be made at a larger scale.
During the LCTH microhabitat study conducted in the San Cristobal Valley (Blackman at
el. 2010), we did not collect some measurements (e.g., distance to nearest tree and wash)
18
that were collected in this study. Numbers of trees within plots, nearest tree distance 10-
50 m, and nearest wash distance 0-10 m were all component variables of the most
parsimonious LCTH occupancy model. The number of trees covariate produced a
parameter estimate indicating a negative relationship, as did nearest tree distance 10-50 m
(Table 4). These results could be explained by the general paucity of trees throughout
LCTH habitat and all “nearest distance to tree” data was collected for points where no
trees were found within the plot. Trees or arborescent structures (e.g., large shrubs) are
important for LCTH nesting and that our model suggests a negative relationship between
trees and LCTH occurrence is surprising. However, our results could be a function of
scale in that LCTH select for trees at the scale of their home range and trees are usually
very sparse throughout LCTH habitat in general. Thus, LCTH apparently select areas
with low tree density, but not completely devoid of trees.
Several tree species are important to LCTH nesting, including crucifixion thorn and
mesquite hummocks (mesquite hummocks act as tree islands within a sparsely vegetated
landscape). Only three nests were documented in 2011. Finding LCTH nests is time
consuming and was not within the scope of this project. Each nest was in a different
plant species. Other studies have documented nesting in large shrubs and even
abandoned buildings and vehicles (Sheppard 1970). LCTH nest-site selection may be
more driven by vegetation structure than plant taxonomy or diversity.
LCTH detections did not appear to correlate with NRCS soil map units or associations.
Correspondingly, all modeled soil types ranked low compared to other covariates in
occupancy modeling. However, LCTH could select habitat that is fundamentally
described by soil data, but patterns may not have been observable at the scale of this
study. Additionally, many soil map units and associations superficially contain very
similar characteristics (e.g., soil substrate and vegetation composition) within LCTH
habitat. Thus, soil attributes may be too fine a scale for modeling relationships regarding
LCTH occurrence. However, our data does reveal soil map units and associations that
definitively contain LCTH detections (Tables 5 and 6); a variable that could be useful in
its own right.
Occupancy models are useful tools, especially if relationships between habitat attributes
and response variables can be discovered (Kroll et al. 2007, Mackenzie 2006). This
study presented occupancy models in the context of habitat variables and how they
pertain to LCTH occupancy and detection probabilities. However, several other variables
(ultimate factors) influence LCTH distribution including: food availability, inter and
intra-specific resource competition, depredation, and inclement weather (e.g., infrequent
storms and consecutive hard freezes). Habitat variables (proximate factors) may often act
as a surrogate for other factors, such as food availability, that affect bird distribution.
However, it is difficult to test the relationship between proximate and ultimate factors and
this study modeled only the effects of habitat covariates. Furthermore, it is not possible
to test the influence of urbanization, technology development (e.g., solar and wind energy
projects) and agriculture footprints within the scope of this study as the vast majority of
the study areas were either undeveloped or within restricted access sites.
19
Our results indicate that LCTH have an inherently low detection probability and
demonstrate that we were able to examine the influence of site-specific covariates on
PAO and detection probabilities under the occupancy modeling framework (MacKenzie
et al. 2002). LCTH occupancy and detection probabilities changed with the scale at
which the LCTH occupancy data were analyzed. For example, occupancy and detection
probabilities differed markedly between odd and even broadcast points (when pooled
respectively) and between transect A and transect B. In this study, occupancy modeling
at the scale of the survey plot was most appropriate. We recommend that future
modeling efforts be conducted at the plot scale. However, as it is important to document
as many LCTH detection locations as is feasible, we also recommend that the same
survey protocol that was used in 2011 be implemented in 2012 when the second phase of
this study occurs. Additionally, we recommend that 2012 survey efforts generate a new
set of random plots to be surveyed from those surveyed in 2011. We also recommend
that all plots be surveyed four times, even if this requires less overall survey plots be
conducted. In 2012, we will continue to monitor LCTH with presence/absence data at
survey plots randomly distributed within BMGR and YPG. We will also continue to
gather the same habitat data as in 2011 but will additionally collect information
pertaining to additional landscape variables: plot distances to mountain foothills and
major valley centers, and numbers of proximal washes and trees.
The Le Conte’s thrasher Prediction of Occurrence Model will be updated with 2012
survey location data as it becomes available. This model will also be refined with
landscape-scale geospatial data obtained by fine-scale imagery such as: distance to
mountains and center of the valley; distance to desert pavement; shrub cover at the plot
scale and tree associations at the plot scale. Refining the model will allow land managers
to predict potential LCTH habitat with greater accuracy and overlay other layers such as
military training areas, bombing ranges and areas slated for development footprints (such
as wind and solar array locations). This will allow land managers to print large maps
containing these overlays and disseminate to the appropriate agents on the ground.
This is the first large-scale study to model the occupancy and detection probabilities of
LCTH and provides a benchmark against which future research can be compared. Long-
term research is critical for separating natural from anthropocentric fluctuations in
wildlife populations and occupancy modeling can provide a reliable alternative to more
costly and labor intensive methods for estimating abundance. Despite repeated visits to
the same survey locations, occupancy modeling is not in itself an exclusive monitoring
technique for determining whether the LCTH population is self-sustaining. Attaching
radio-transmitters and tracking LCTH would be useful to elucidate individual movement
data such as home range and breeding territory sizes and territory shifts. To determine a
population’s self-sustainability, it is necessary to gather productivity and survival data
through time; however, these types of data are costly to acquire (Henneman and
Andersen 2009). High occupancy rates through time at repeated survey locations can
indicate a relatively healthy population within areas spared from high anthropogenic
impacts.
20
MANAGEMENT AND RESEARCH PRIORITIES
The US Census Bureau projected that Arizona would add 5.6 million people by 2030,
making it the 10th most populated state in the country and ranking in the top five fastest-
growing states. Within BMGR and YPG are large expanses of relatively undisturbed
Sonoran Desert, mostly of the Lower Colorado River Subdivision. The importance of
these un-fragmented areas to LCTH and many other lowland desert species will continue
to increase as the landscape surrounding these DoD installations is rapidly developed for
agriculture, industries such as alternative energy (solar), and urban expansion.
The DoD installations of southern Arizona, along with the US Fish and Wildlife Service,
AZ Game and Fish Department, Bureau of Land Management, National Park Service,
Bat Conservation International and Sonoran Joint Venture, are partners in the Sonoran
Desert Conservation Partnership Team. In 2007 this team produced DoD Legacy
Species-at-Risk documents that, based on the information available at the time,
synthesized the ecology and management recommendations for the species of concern
shared by the three DoD ranges of southwestern Arizona. Le Conte’s thrasher is the only
bird among these shared Species-at-Risk. This study addressed the following
recommended management and research priorities for LCTH made by the Department of
Defense Species-at-Risk project
Collect data on LCTH distribution in order to evaluate this species’ distribution in
relation to military training activities and potential threats. This study collected the
first season of LCTH occupancy data used to predict occurrence patterns
(incorporating a detection probability) and also facilitated setting a benchmark for
comparison with future surveys.
Evaluate effects of habitat conditions and land use on LCTH populations to
develop better understanding of their distribution and support development of
appropriate management actions. An objective of this study was to describe
essential habitat components for LCTH; areas containing softer substrate and
containing minimal or no gravel composition adequate for digging/ground
gleaning, prominence of washes, and sparse tree composition all comprise LCTH
habitat.
Concentrate training and development activities away from areas with current or
historic records of Le Conte’s thrashers. In addition, evaluate potential impacts to
the local viability of thrashers, including habitat loss and fragmentation, when
developing new training areas. This approach should reduce disturbance to
important areas for LCTH and other species while reducing overall fragmentation
of wildlife habitat. This study has provided more detailed locations of LCTH
habitat within the BMGR and YPG and highlights important components of LCTH
habitat that can be extrapolated to other areas. Thus, areas potentially suitable to
LCTH occurrence can be more easily predicted across this species distribution and
particularly where potential military training-related impacts are planned. This
study also developed a LCTH Prediction of Occurrence Model that can be
21
combined with geospatial data of military training activities, bombing ranges and
areas slated for development (e.g., solar and wind arrays) to most effectively
predict the potential impacts that these actions may have on LCTH.
Create or maintain OHV closure to Le Conte’s thrasher breeding areas. The
borderlands region of the U.S. experiences a multitude of OHV disturbance from
illegal activity and border patrols. LCTH surveyors noted that OHV footprints
were ubiquitous throughout the study area and will continue to be difficult to
police. Most survey locations contained evidence of OHV footprints to some
degree. While LCTH persisted in many of these areas, the borderlands region
receives regular traffic from OHVs and reducing this traffic falls under the auspices
of the Department of Homeland Security. Determining the impacts of OHV use on
LCTH occurrence is beyond the scope of this project and would be difficult and
costly to achieve. We will attempt to include OHV footprints as a covariate during
year 2 analyses.
22
The following research priorities would address knowledge gaps with respect to Le
Conte’s thrasher ecology and would improve our ability to proactively manage its habitat:
Evaluate disturbance threshold of OHV to Le Conte’s thrasher populations in the
US and Mexico. OHV footprints are widespread throughout the borderlands region
and it is difficult to correlate the impacts of OHV traffic on LCTH.
Compare the habitat that LCTH are using versus what is available to them. This
can be accomplished by measuring habitat variables at plots within Le Conte’s
territories in conjunction with measuring habitat variables at random plots.
Other potential disturbances to LCTH are expected to increase including urban and
agricultural development, wind and solar power. In the face of these potential
threats, it is important to investigate the thresholds to which LCTH respond
negatively to these disturbances. The refined LCTH PO Model (after year 2) will
be a powerful tool for land managers to predict potential areas containing LCTH
are most sensitive to disturbance. Describing actual threshold values that LCTH
respond to in the context of development would require extensive tracking efforts
(i.e., radio-telemetry) in areas outside of DoD responsibility.
Initiate or continue monitoring the expansion of invasive plant species and their
impact on Le Conte’s thrasher populations in the US and Mexico.
Develop and implement integrated management strategies to reduce wildfire fuel
loads and further spread of invasive species. Evaluate effects of invasive species
management on LCTH populations in US and Mexico.
23
REFERENCES
Bailey, L. L., T. R. Simons, and K. H. Pollock. 2004. Estimating site occupancy and
species detection probability parameters for terrestrial salamanders. Ecological
Applications 14:692-702.
Barry M. Goldwater Range (BMGR). 2007a. Integrated Natural Resources Management
Plan (INRMP) (2007 – 2011), Barry M. Goldwater Range. Prepared by U.S.
Department of the Air Force, Luke Air Force Base and U.S. Department of the
Navy, Marine Corps Air Station Yuma. March 2007.
Barry M. Goldwater Range (BMGR). 2007b. Inventory and Management Plan, Barry
M. Goldwater Range-East. Science Applications International Corp (SAIC),
Tucson, AZ. ACES Project No. NURD71937. August 2007.
Blackman, S. T., S. F. Lowery and J. E. Diamond. 2010. Le Conte’s Thrasher broadcast
survey and habitat measurement, Barry M. Goldwater Range – East. Research
Branch, Arizona Game and Fish Department.
Brown, D. E. 1994. Biotic communities: Southwestern United States and Northwestern
Mexico. University of Utah Press. Salt Lake City, Utah.
Buckland S. T., D. R. Anderson, K. P. Burnham and J. L. Laake. 2001. Distance
Sampling: Estimating Abundance of Biological Populations. Chapman and Hall,
London. 446 pp.
Burnham, K. P. and D. R. Anderson. 2002. Model selection and multimodel inference: a
practical information-theoretic approach. Springer-Verlag. New York, New York.
CalPIF (California Partners in Flight). 2006. Version 1.0. The Desert Bird Conservation
Plan: a Strategy for Protecting and Managing Desert Habitats and Associated Birds
in the Mojave and Colorado Deserts. California Partners in Flight. Available:
http://www.prbo.org/calpif /plans.html
Coachella Multi-species Habitat Conservation Plan (CMHCP). 2007. Available:
http://www.cvmshcp.org.
Collier, B. A., M. L. Morrison, S. L. Farrell, A. J. Campomizzi, J. A. Butcher, K. B.
Hays, D. I. MacKenzie and R. N. Wilkins. 2010. Monitoring golden-cheeked
warblers on private lands in Texas. Journal of Wildlife Management 74(1): 140-
147.
Corman, T. E. and Wise-Gervais, C. 2005. Arizona Breeding Bird Atlas. University of
New Mexico Press, Albuquerque, NM.
Hansen, C. P., J. J. Millspaugh and M. A. Rumble. 2011. Occupancy modeling of ruffed
grouse in the Black Hills National Forest. Journal of Wildlife Management 75(1):
71-77.
Henneman, C. and D. E. Andersen. 2009. Occupancy models of nesting-season habitat
associations of red-shouldered hawks in central Minnesota. Journal of Wildlife
Management 73(8): 1316-1324.
Hines, J. E. 2010. Program PRESENCE. U.S. Geological Survey, Patuxent Wildlife
Research Center, Laurel, Maryland.
Kroll, A. J., S. D. Duke, D. E. Runde, E B. Arnett and K. A. Austin. 2007. Modeling
habitat occupancy of orange-crowned warblers in managed forests of Oregon and
Washington, USA. Journal of Wildlife Management 71(4): 1089-1097.
24
MacKenzie, D. I, J. D. Nichols, G. B. Lachman, J. A. Royle, and C. A. Langtimm. 2002.
Estimating site occupancy rates when detection probabilities are less than one.
Ecology 82:2248-2255.
MacKenzie, D. I., J. D. Nichols, J. E. Hines, M. G. Knutson, and A. D. Franklin. 2003.
Estimating site occupancy, colonization and local extinction probabilities when a
species is not detected with certainty. Ecology 84:2200-2207.
MacKenzie, D. I., and L. L. Bailey. 2004. Assessing the fit of site occupancy models.
Journal of Agricultural, Biological, and Environmental Statistics 9:300–318.
MacKenzie, D. I., and J. D. Nichols. 2004. Occupancy as a surrogate for abundance
estimation. Animal Biodiversity and Conservation 27:461-467.
MacKenzie, D. I., and J. A. Royle. 2005. Designing occupancy studies: general advice
and allocating survey effort. Journal of Applied Ecology 42:1105-1114.
MacKenzie, D. I. 2006. Modeling the probability of resource use: the effect of, and
dealing with, detecting a species imperfectly. Journal of Wildlife Management
70:367-374.
MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, and J. E. Hines.
2006. Occupancy estimation and modeling: inferring patterns and dynamics of
species occurrence. Academic Press, San Diego, California, USA.
Latta, M. J., C. J. Beardmore and T. E. Corman. 1999. Arizona Partners in Flight Bird
Conservation Plan. Version 1.0. Nongame and Endangered Wildlife Program
Technical Report 142. Arizona Game and Fish Department, Phoenix, Arizona.
Purcell, K. L., S. R. Mori and M. K. Chase. 2005. Design considerations for examining
trends in avian abundance using point counts: examples from oak woodlands.
Condor 107: 305–320.
Sheppard, J. M. 1970. A study of the Le Conte’s Thrasher (Toxostoma lecontei).
California Birds 1: 85-94.
Sheppard, J. M. 1996. Le Conte’s Thrasher (Toxostoma lecontei). In Birds of North
America 230 (A. Poole and F. Gills, Eds.). Philadelphia: Academy of Natural
Sciences; Washington, DC: American Ornithologists’ Union.
Tyre, A. J., B. Tenhumberg, S. A. Field, D. Niejalke, K. Parris, and H. P. Possingham.
2003. Improving precision and reducing bias in biological surveys by estimating
false negative error rates in presence-absence data. Ecological Applications
13:1790-1801.
Villarreal, M.L., C. van Riper III, R.E. Lovich, R.L. Palmer, T.Nauman, S.E. Studd, S.
Drake, A.S. Rosenberg, J. Malusa, and R.L. Pearce. 2011. An Inventory and
Monitoring Plan for a Sonoran Desert Ecosystem: Barry M. Goldwater Range-
West. U.S. Geological Survey Open-File Report 2011-1232.
25
Appendix 1. Randomly generated centers of Le Conte’s thrasher survey plots
Plot
ID Range
LCTH Detection
Locations
Easting (NAD
83)
Northing (NAD
83)
1 MCAS 12 231649 3615117
2 MCAS 8 774657 3607203
3 BMGRE 3 291232 3623229
5 MCAS 8 225385 3609340
6 BMGRE 12 272547 3620943
7 MCAS 1 777233 3599482
8 MCAS 4 735842 3607476
9 BMGRE 4 261144 3607159
10 YPG 0 753796 3667076
11 MCAS 4 249740 3604883
12 MCAS 1 772003 3579622
13 MCAS 4 241355 3609949
14 YPG 6 245033 3657724
15 YPG 0 744247 3655608
16 MCAS 1 250544 3595417
17 BMGRE 5 271492 3593667
18 MCAS 0 781063 3596174
19 BMGRE 3 249249 3622850
20 YPG 1 248916 3650748
21 BMGRE 6 256996 3619594
22 YPG 0 247397 3676847
23 MCAS 0 775543 3588063
24 BMGRE 0 293373 3631604
25 MCAS 1 781627 3591141
26 MCAS 5 771174 3600639
28 BMGRE 5 274502 3598025
29 BMGRE 0 252945 3616304
30 MCAS 1 236413 3593932
31 MCAS 0 767069 3584002
32 YPG 0 757467 3654720
33 MCAS 2 762474 3607583
34 BMGRE 1 333898 3609304
35 MCAS 7 774395 3593091
36 MCAS 5 230508 3601506
37 BMGRE 8 268101 3614367
38 BMGRE 0 249970 3619363
40 BMGRE 6 266659 3598273
22-2 YPG 0 250663 3681422
41 BMGRE 0 328729 3603609
43 YPG 5 247912 3653930